CFD continuation, bifurcation detection
Posted: 20 Oct 2014, 16:49
Hello,
we developped an Elmer solver in a dynamic library for continuation of steady flow solution and bifurcation detection.
It's based on the perturbation method called Asymptotic Numerical Method.
Those tools are allready improved and published.
We decided to try Elmer FEM to get 3D and mutli-physics flows informations.
I'll try to put a presentation of what we did in this Elmer library, and propose a "clean" version of the code to perform continuation and bifurcation detection.
I'd like to know what is the best way to get advices on the way we coded this technique, and further questions we have.
Should i put the code in here, or in a more specific mail?
For example :
+ We had to create a new Dirichlet Boundary condition routine, in order to modify values at nodal point in a simple way.
=> it might exist a smarter way to code what we did, respecting an "Elmer way of coding".
+ Once a bifurcation is detected via specific indicators, branch switching requires an "augmented linear system" to be solved, because the tangent operator is singular at the bifurcation:
=> Do you know how to add temporarily 1 or 2 equation to the "solver % Matrix" object, and perform a linear solve of this augmented system?
=> Would it be easier to create a new temporary solver with the correct given number of equation to be solved. And how to do that?
Thank you for your advices
we developped an Elmer solver in a dynamic library for continuation of steady flow solution and bifurcation detection.
It's based on the perturbation method called Asymptotic Numerical Method.
Those tools are allready improved and published.
We decided to try Elmer FEM to get 3D and mutli-physics flows informations.
I'll try to put a presentation of what we did in this Elmer library, and propose a "clean" version of the code to perform continuation and bifurcation detection.
I'd like to know what is the best way to get advices on the way we coded this technique, and further questions we have.
Should i put the code in here, or in a more specific mail?
For example :
+ We had to create a new Dirichlet Boundary condition routine, in order to modify values at nodal point in a simple way.
=> it might exist a smarter way to code what we did, respecting an "Elmer way of coding".
+ Once a bifurcation is detected via specific indicators, branch switching requires an "augmented linear system" to be solved, because the tangent operator is singular at the bifurcation:
=> Do you know how to add temporarily 1 or 2 equation to the "solver % Matrix" object, and perform a linear solve of this augmented system?
=> Would it be easier to create a new temporary solver with the correct given number of equation to be solved. And how to do that?
Thank you for your advices